Image Tools guide
Prepare an image for the web, a form or a profile without guessing
Image preparation becomes much easier when you separate three things that are often called “size”: the area you keep, the pixel dimensions you deliver, and the number of bytes in the encoded file. Fix them in that order when the destination requires all three.
Translate the destination rule into an image requirement
A requirement such as “1200 × 1200 maximum” describes pixels. “Under 2 MB” describes encoded bytes. “Square profile photo” describes aspect ratio or composition. A single image can satisfy two and still fail the third, so record each constraint before editing.
Format can be a fourth constraint. A portal may accept JPG but reject PNG, or a design workflow may require transparency that JPG cannot store. Do not compress repeatedly before checking whether the receiving system is rejecting dimensions, bytes, shape or format.
| Action | What it changes | What it cannot promise |
|---|---|---|
| Crop | Which region remains and possibly the aspect ratio | A specific pixel width or file size |
| Resize | Pixel width and height | A specific KB/MB result |
| Compress | Encoded representation and sometimes visual quality | A different composition |
| Change format | Encoding capabilities and compatibility | Restoration of detail already lost |
Crop before shrinking when composition is part of the requirement
If an avatar must be square, choose the square composition while the source still has all its detail. Then resize that crop to the required dimensions. Compressing a large uncropped photo first can throw away detail in areas you later discard.
Cropping is not resizing. It removes pixels outside the selected rectangle. For faces, documents and product labels, inspect the edges deliberately: a mathematically correct square is still a bad crop if it cuts off the information the image exists to show.
- Input
- A 2400 × 1600 photo needed as a 1200 × 1200 profile image
- Method
- Choose a square crop around the subject, then resize that square to 1200 × 1200
- Result
- Correct shape without stretching the original 3:2 photograph
Resize to the dimensions people will actually use
A 4000 × 3000 image contains twelve million pixels. At 2000 × 1500 it contains three million. That is a 75% reduction in pixel count, but it is not a promise of 75% fewer file bytes because encoders respond differently to detail, noise, transparency and repeated patterns.
Keep the aspect ratio locked unless distortion is intentional. A 2400 × 1600 photograph resized to 1200 pixels wide naturally becomes 1200 × 800. Forcing it to 1200 × 1200 without cropping changes the shape of faces, circles and other recognizable objects.
Why a PNG can stay surprisingly large
PNG stores processed pixels without lossy image compression. Flat colors, repeated areas and transparency can compress efficiently, while camera noise, texture and gradients are harder to represent compactly. Two PNGs with identical dimensions can therefore have very different byte sizes.
A large transparent canvas also still has dimensions and decoding cost even when much of it looks empty. SnakTool's PNG compression is lossless for the processed pixels and may have little to remove from an already efficient file. It does not expose palette quantization as a separate optimization strategy.
When format choice is a better fix than more compression
A detailed photograph with no transparency may be more practical as JPG or WebP than PNG. A screenshot, diagram or logo with sharp edges can favor PNG when preserving those edges matters more than minimum bytes. If transparency is required, do not convert blindly to JPG; decide what background should replace the transparent areas.
Changing JPG to PNG does not restore detail lost in an earlier JPEG encode. Repeated lossy saves can add damage. Keep the original and make destination-specific copies rather than using the latest compressed copy as the source for every future version.
| Image situation | Useful first move | Inspect |
|---|---|---|
| Large photograph, no transparency | Resize, then compare JPG/WebP | Faces, texture and actual bytes |
| Transparent logo | Crop unused canvas, resize if oversized, keep alpha-capable format | Edges on light and dark backgrounds |
| Sharp screenshot with small text | Remove unused area and reduce dimensions carefully | Text at normal viewing size |
| Archival or master image | Keep original unchanged | Create a separate delivery copy |
Compress last, after the pixels you do not need are gone
Once composition and dimensions are correct, compare compression only if the encoded file still exceeds the destination limit. This gives the encoder fewer unnecessary pixels to store and lets you judge quality at the actual delivery dimensions.
A smaller file is not automatically a better result. Check text, faces, product labels, gradients and high-contrast edges at the intended display size. If a strict byte target forces unacceptable damage, reconsider dimensions or the accepted format instead of repeatedly lowering quality.
- Write down shape, pixel, byte and format requirements.
- Crop to the intended composition.
- Correct orientation before measuring final dimensions.
- Resize while preserving the intended aspect ratio.
- Choose a compatible format with the required transparency behavior.
- Compress only as much as needed, then inspect the downloaded result.
Keep one clean source through the whole workflow
Rotation, crop, resize, format conversion and compression are separate transformations. A quarter-turn swaps width and height; a crop removes information; lossy compression can remove detail. Performing each new experiment from an already degraded copy makes comparisons harder.
Keep the original until the destination accepts the final image. That also lets you produce a different version later for a larger screen, another aspect ratio or a transparency-capable workflow without trying to reconstruct information that was discarded.